Local Menu
Media Technologies Research Topics
Unified approach to automatic speech recognition and knowledge integration
Explanation
Integrating grammatical or domain knowledge into speech recognition may lead to dramatic improvements in recognition performance. This programme is aimed at building a framework to unify knowledge and speech recognition in a statistical manner. The research project will include creating novel acoustic models and performance evaluation. ICASSP and Interspeech cover the field of the research topic.
Knowledge and Skills required
Candidates should possess expertise on speech recognition or statistical machine translation and C, C++ programming skills.
Emotional speech synthesis
Explanation
The aim of this research is to provide expressive or emotional features to synthetic speech. The research project will include algorithm development and subjective evaluation of the algorithm. The algorithm will cover spectrum modelling, source excitation models for emotional speech, training of model parameters and model adaptation to new voices. Toshiba has in depth expertise on speech synthesis. ICASSP and Interspeech cover the field of the research topic.
Knowledge and Skills required
Candidates should possess expertise on speech synthesis, statistical machine learning and C, C++ programming skills.
Machine vision for object recognition
Explanation
The subject of this project is to develop a recognition technology for 3-D objects with the aim of producing useful vision applications, such as video image retrieval. The research theme may include: feature detection, object classification, segmentation, pose estimation and effective search. The field of research is expected to cover that represented by the ICCV and CVPR conferences.
Knowledge and Skills required
Candidates are required to possess expertise in computer vision and/or machine learning together with programming skills sufficient to formulate fundamental experiments.
Dialogue interface for robots
Explanation
This project pursues the possibilities for enabling a robot to become a conversational interface and is aimed at building the necessary elemental technologies. The approach relies intensively on nonverbal information and conversations carried out primarily in Japanese have been studied. This research will look for insights on the applicability of this approach to European languages.
Knowledge and Skills required
Candidates should possess software skills in either computer vision, speech recognition or sensor signal processing. They should also have experience in research on communication robots, human-machine interactions or related areas.
